Gerald Ajam

What dashboards cannot tell you

A dashboard looks like a view of the whole class. Its edges are the edges of the software. Paper, conversation and home, the start of the unit and the months after it all sit outside the frame, and a well-made chart makes the frame easy to forget.

Status Position

Published November 2026

All notes

In short

A dashboard shows only what the platform could see, so the questions that matter most about learning have to be answered from somewhere else.

The shift it turns on

The usual reading

The dashboard as a window

Open it, and see how learning is going.

The reading it moves to

The dashboard as a frame

It shows what the platform saw. The rest has to be asked for.

01The situation

Everything the software saw

Dashboards are where learning data meets the people who decide things. A teacher opens one between lessons, a school leader before a meeting and a product team every morning. Each sees clean charts, sorted lists and coloured flags, and each reasonably assumes the screen is showing the situation.

It is showing the situation as the software saw it. A platform sees what happens inside it: the clicks, the submissions and the answers to the items it set. It does not see the worksheet, the discussion, the tutor at home or the sibling sharing the device. A child’s log records the household and the timetable as much as the learning.

The evidence that learning analytics, dashboards included, improves learning is thin. Viberg and colleagues (2018) reviewed 252 papers on learning analytics in higher education. Nine per cent showed evidence of improved learning outcomes.

02The tension

What sits outside the frame?

What if the questions that matter most about learning are the ones a dashboard is least able to answer? Four come up in nearly every review. Why did it happen? Where were learners before? Will it last? What would have happened anyway?

Each falls outside the frame for a structural reason, and better charting cannot fix a structural reason. A dashboard can describe what learners did on the platform this week. It can help a teacher narrow down why. It has no access to the starting point unless someone measured it, none to next month and none to the world in which the product was never used.

A dashboard has no access to the starting point unless someone measured it, none to next month and none to the world in which the product was never used.

03The move

Four questions to take elsewhere

The move is to name what the screen cannot answer, and to decide in advance where each answer will come from.

  1. Why: ask the learners

    A long time on an item could mean struggle, care or a tab left open. A low score could mean a misconception, a badly worded item or a shared slip. The dashboard shows the pattern. The reason is in what learners wrote and said, so read ten of the wrong answers and ask three learners what they were thinking before deciding anything.

  2. Before: measure the start

    Mastery shown at the end of a unit may have been there at the beginning. Unless the platform set a starting check, the dashboard is reporting a state, and the reader is left to treat it as a change.

  3. Later: schedule the return

    The signals that arrive in time to act on are the ones furthest from learning. Clicks arrive at once and quiz scores within days. The outcomes that matter arrive months or years later. A dashboard lives at the fast end, and whether this week’s mastery survives to next term is a question for a check that has not happened yet.

  4. Anyway: find the comparison

    An early-warning system at an American university was widely cited for raising retention, because students who took more courses using it stayed on at higher rates. Caulfield (2013) pointed out the reverse explanation. Students who stay longer take more courses, so they have more chances to take one that uses the system. The data showed who persisted. Showing what the system changed needed a comparison the dashboard was never going to contain.

04Where it lands

A frame that marks its own edges

The first change is in how a dashboard is read. Before opening one, write the question down. “Who has not started?” and “Where did many learners go wrong?” are questions a platform answers well. “Is this working?” is one it cannot answer alone, and reading the screen for it produces worry and few decisions.

The second is in how one is designed. A good dashboard marks its own edges: what it measured and what it left out, when it was last checked against a real outcome and which learners it describes poorly, such as those who work mostly offline. I would rather use a modest dashboard that says where it ends than an impressive one that seems to cover everything.

The third is a rule for meetings. A claim about learning made from a dashboard comes with the source of its why, its before, its later and its anyway. Where one is missing, the claim is reported as a description of activity, which is what it is.

The point

What transfers

The reason this page exists, rather than the story that produced it.

  1. A dashboard shows what the platform saw. The worksheet, the conversation and the home are outside the frame.
  2. Four questions sit outside every dashboard: why it happened, where learners started, whether it lasted and what would have happened anyway.
  3. Decide in advance where each answer will come from: the learners’ own words, a starting check, a later check and a comparison.
  4. Write the question before opening the screen. A dashboard read without one produces worry and few decisions.
  5. Build dashboards that mark their own edges: what was measured, what was left out and whom the data describes poorly.
Singapore